CN107622277A - A kind of complex-curved defect classification method based on Bayes classifier - Google Patents

A kind of complex-curved defect classification method based on Bayes classifier Download PDF

Info

Publication number
CN107622277A
CN107622277A CN201710748823.6A CN201710748823A CN107622277A CN 107622277 A CN107622277 A CN 107622277A CN 201710748823 A CN201710748823 A CN 201710748823A CN 107622277 A CN107622277 A CN 107622277A
Authority
CN
China
Prior art keywords
image
defect
sample
complex
bayes classifier
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201710748823.6A
Other languages
Chinese (zh)
Other versions
CN107622277B (en
Inventor
陈振州
张平
张美杰
尹志锋
汪南辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201710748823.6A priority Critical patent/CN107622277B/en
Publication of CN107622277A publication Critical patent/CN107622277A/en
Application granted granted Critical
Publication of CN107622277B publication Critical patent/CN107622277B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The present invention relates to a kind of complex-curved defect classification method based on Bayes classifier, comprise the following steps:S1, collection surface chart picture；S2, the image collected is pre-processed；The ROI region of defect part is included in S3, extraction image, obtains imageS4, to imageSurface fitting is carried out, obtains fitted figureS5, by imageWith fitted figureMake difference processing, obtain defect information present in ROI region；S6, prominent defect part；S7, structure Bayes classifier；S8, on-line checking are simultaneously classified.The present invention can complete detection and the classification of surface defect on the basis of workpiece surface complexity is considered；Meanwhile accuracy of detection is high, has preferable practicality.

Description

A kind of complex-curved defect classification method based on Bayes classifier
Technical field
The present invention relates to the technical field of defects detection, more particularly to it is a kind of based on the complex-curved of Bayes classifier Defect classification method.
Background technology
Vision detection technology is widely applied in surface defect automatic detection field.But traditional surface defects detection Method is then related to few for complex-curved defects detection mainly in the object of regular planar texture-free.It is complicated bent Planar defect is detected due to its complicated geometrical properties, causes image procossing larger difficulty to be present.
Have related article before and studied application of the computer vision in terms of curved surface, such as document Automated Surface inspection for directional textures [J] .Image＆Vision Computing and document A novel internal thread defect auto-inspection system[J].International Journal of Advanced Manufacturing Technology.But unfortunately, the Detection results of above-mentioned document methods described Both limited by the complexity on surface to be checked, the defects of being not suitable for complex geometry curved surface, is detected.
How on the basis of workpiece surface complexity is considered, detection and the classification of complex geometry surface defect are completed, into To produce, the producer of complex geometry curved surface object (such as leather, curve surface work pieces, plastic bottle, automobile engine cover) is urgently to be resolved hurrily to be asked Topic.
The content of the invention
Go out complex geometry curved surface it is an object of the invention to overcome the deficiencies of the prior art and provide a kind of energy effective detection to lack Sunken and accuracy of detection is high, the good complex-curved defect classification method based on Bayes classifier of practicality.
To achieve the above object, technical scheme provided by the present invention is：Comprise the following steps：
S1, collection surface chart picture；
S2, the image collected is pre-processed；
The ROI region of defect part is included in S3, extraction image, obtains image
S4, to imageSurface fitting is carried out, obtains fitted figure
S5, by imageWith fitted figureMake difference processing, obtain defect information present in ROI region；
S6, prominent defect part；
S7, structure Bayes classifier；
S8, on-line checking are simultaneously classified.
Further, step S2 image preprocessings include image gray processing and medium filtering.
Further, it is as follows to extract the step of ROI region that defect part is included in image by step S3：
S3-1, using iterative method Calculation Estimation threshold value T', be specially：
S3-1-1, gradation of image average T is asked for by grey level histogram, be designated as initial threshold；
S3-1-2, use T segmentation figure pictures；Image is divided into all pixels set G1 of the gray value more than T and gray value is small In T all pixels set G2；
S3-1-3, calculating G1 and the pixel in G2 set average gray value μ 1 and μ 2；
S3-1-4, using 1/2 (μ 1+ μ 2) the threshold value T's new as one；
S3-1-5, repeat step S3-1-2~S3-1-4, until the difference of two T' in subsequent iteration is than previously given Untill parameter value is small；
S3-2, Utilization assessment threshold value T' extract the ROI region for including defect part：
The all pixels point in image is analyzed, if pixel gray value is less than threshold value T' in image to be checked, regards it as background Point or non-detection region point, reject the pixel；Conversely, being considered as in detection object the point in detection zone, retain the pixel； Image to be detected A (Xi,Yi,Zi) extraction ROI after obtain the new images containing only region to be detectedWherein, Xi,Yi For pixel point coordinates, ZiRepresent the pixel gray value.
Further, step S4 surface fittings are specially：
Using least square method structure fitting function Z=F (X, Y)；By imageThe coordinate of middle pixel Xi,YiWith gray value ZiAs fitting function Zi=f (Xi,Yi) sample data, obtain fitting function Z=F (X, Y) with it is to be detected Region fitted figure
The solution of least square fitting is：For the sample point (X on testing imagei,Yi,Zi), look for one in function space Individual function Z=F*(X, Y), make error sum of squares minimum, i.e. the error of fitting result and sample data is minimum, is optimal plan Close；Wherein, using Euclid norms | | δ | |2As error metrics standard.
Further, in step S6, removed by the difference result to step S5 using rim detection and morphology operations Image border and tiny flaw part, prominent defect part.
Further, step S7 builds comprising the following steps that for Bayes classifier：
S7-1, structure database model：
Including positive sample and negative sample, positive sample is zero defect image, and negative sample is various types of defect images.
S7-2, sample characteristics extraction：
Sample characteristics mainly include gray feature and geometric properties；Because the gray feature of positive sample, geometric properties have Very strong regularity is obvious with negative sample difference, it is easy to distinguish；Specifically include following steps：
S7-2-1：Positive sample and the gray feature information of negative sample are obtained by grey level histogram, specifically include gray average, Gray scale intermediate value, minimum or maximum gray scale, the pixel count more than or less than setting；Wherein, gray average refers to institute in region There is the average value of pixel；Gray scale intermediate value refers to the sequence intermediate value of all pixels in region；By these eigenvalue clusters into a gray scale Characteristic vector group, is designated as H, the evaluation criterion as gray feature；
S7-2-2：Analyzed to obtain positive sample and the geometric properties information of negative sample by Blob, specifically include area, circle Degree, rectangular degree, girth and length-width ratio.By these eigenvalue clusters into a geometric properties Vector Groups, G is designated as, as geometric properties Evaluation criterion.
S7-3, structure Bayes classifier, are comprised the following steps that：
S7-3-1, prepare training sample：
There to be same characteristic attribute, same category of positive negative sample is divided into same set；Wherein, characteristic attribute is The gray feature Vector Groups H and geometric properties Vector Groups G that step S7-2 is drawn；
S7-3-2, training grader：
Calculate three prior probabilities, respectively P (D), P (f | D), P (D | f), generate grader；And then by Bayes' theorem The posterior probability of all kinds of positive negative samples is calculated, realizes classification.
Bayes' theorem calculation formula is：
Wherein, P (f) represents to assume defect f prior probability, and f is a division for assuming set；P (D) represents positive and negative instruction Practice the prior probability of sample；The probability distribution of each characteristic value during P (D | f) is represented per a kind of defect sample, and P (f | D) represent through shellfish The posterior probability for all kinds of defects of hypothesis that this classifier calculated of leaf obtains；
S7-3-3, testing classification performance：
It is corresponding using the posterior probability of the step S7-3-2 all kinds of defects of classifier calculated for training to obtain, maximum a posteriori probability Classification exported as classification results, whether by output result and standard results contrast verification, it is accurate to judge to classify.
Compared with prior art, this programme advantage is as follows：
This programme on the basis of workpiece surface complexity is considered, can complete the detection of complex geometry surface defect and divide Class；It is more the shortcomings that being very difficult to train compared to SVMs and be difficult to explain its inherent laws in the selection of grader Layer perceptron needs mass data to be trained the very high hardware configuration model of training requirement to be in " black box state ", it is difficult to understands The shortcomings that selection of internal mechanism member parameter and network topology is difficult, Bayes classifier have it is quick, be easy to training, training effect Well, the high advantage of discrimination；Meanwhile accuracy of detection is high, has preferable practicality.
Brief description of the drawings
Fig. 1 is a kind of flow chart of the complex-curved defect classification method based on Bayes classifier of the present invention.
Embodiment
With reference to specific embodiment, the invention will be further described：
Referring to shown in accompanying drawing 1, a kind of complex-curved defect classification side based on Bayes classifier described in the present embodiment Method, comprise the following steps：
S1, collection surface chart picture；
S2, image gray processing and medium filtering are carried out to the image collected；
The ROI region of defect part is included in S3, extraction image, obtains imageDetailed process is：
S3-1, using iterative method Calculation Estimation threshold value T'：
S3-1-1, gradation of image average T is asked for by grey level histogram, be designated as initial threshold；
S3-1-2, use T segmentation figure pictures；Image is divided into all pixels set G1 of the gray value more than T and gray value is small In T all pixels set G2；
S3-1-3, calculating G1 and the pixel in G2 set average gray value μ 1 and μ 2；
S3-1-4, using 1/2 (μ 1+ μ 2) the threshold value T's new as one；
S3-1-5, repeat step S3-1-2~S3-1-4, until the difference of two T' in subsequent iteration is than previously given Untill parameter value is small；
S3-2, Utilization assessment threshold value T' extract the ROI region for including defect part：
The all pixels point in image is analyzed, if pixel gray value is less than threshold value T' in image to be checked, regards it as background Point or non-detection region point, reject the pixel；Conversely, being considered as in detection object the point in detection zone, retain the pixel； Image to be detected A (Xi,Yi,Zi) extraction ROI after obtain the new images containing only region to be detectedWherein, Xi,Yi For pixel point coordinates, ZiRepresent the pixel gray value.
S4, to imageSurface fitting is carried out, obtains fitted figure
Using least square method structure fitting function Z=F (X, Y)；By imageThe coordinate of middle pixel Xi,YiWith gray value ZiAs fitting function Zi=f (Xi,Yi) sample data, obtain fitting function Z=F (X, Y) with it is to be detected Region fitted figure
The solution of least square fitting is：For the sample point (X on testing imagei,Yi,Zi), look for one in function space Individual function Z=F*(X, Y), make error sum of squares minimum, i.e. the error of fitting result and sample data is minimum, is optimal plan Close；Wherein, using Euclid norms | | δ | |2As error metrics standard.
S5, to ROI region imageWith fitted figureMake difference processing, obtain ROI region In defect information that may be present.
Calculation formula is：
S6, by removing image border and the tiny flaw using rim detection and morphology operations to step S5 difference result Defect part, prominent defect part.
S7, structure Bayes classifier, are comprised the following steps that：
S7-1, structure database model；
S7-2, sample characteristics extraction, process are：
S7-2-1, obtain positive sample and the gray feature value of negative sample by grey level histogram, the eigenvalue cluster drawn is into one Individual gray feature Vector Groups, are designated as H, the evaluation criterion as gray feature；
S7-2-2, analyzed by Blob to obtain the geometrical characteristic of positive sample and negative sample, the eigenvalue cluster drawn is into one Geometric properties Vector Groups, are designated as G, the evaluation criterion as geometric properties.
S7-3, structure Bayes classifier, process are：
S7-3-1, prepare training sample：
There to be same characteristic attribute, same category of positive negative sample is divided into same set；Wherein, characteristic attribute is The gray feature Vector Groups H and geometric properties Vector Groups G that step S7-2 is drawn；
S7-3-2, training grader：
Calculate three prior probabilities, respectively P (D), P (f | D), P (D | f), generate grader；And then by Bayes' theorem The posterior probability of all kinds of positive negative samples is calculated, realizes classification.
Bayes' theorem calculation formula is：
Wherein, P (f) represents to assume defect f prior probability, and f is a division for assuming set；P (D) represents positive and negative instruction Practice the prior probability of sample；The probability distribution of each characteristic value during P (D | f) is represented per a kind of defect sample, and P (f | D) represent through shellfish The posterior probability for all kinds of defects of hypothesis that this classifier calculated of leaf obtains；
S7-3-3, testing classification performance：
It is corresponding using the posterior probability of the step S7-3-2 all kinds of defects of classifier calculated for training to obtain, maximum a posteriori probability Classification exported as classification results, whether by output result and standard results contrast verification, it is accurate to judge to classify.
S8, on-line checking is carried out to product to be measured using the grader of above-mentioned structure and classified.
The present embodiment on the basis of workpiece surface complexity is considered, can complete the detection of complex geometry surface defect and divide Class；Meanwhile accuracy of detection is high, has preferable practicality.
Examples of implementation described above are only the preferred embodiments of the invention, and the implementation model of the present invention is not limited with this Enclose, therefore the change that all shape, principles according to the present invention are made, it all should cover within the scope of the present invention.

Claims (8)

  1. A kind of 1. complex-curved defect classification method based on Bayes classifier, it is characterised in that：Comprise the following steps：
    S1, collection surface chart picture；
    S2, the image collected is pre-processed；
    The ROI region of defect part is included in S3, extraction image, obtains image
    S4, to imageSurface fitting is carried out, obtains fitted figure B；
    S5, by imageWith fitted figureMake difference processing, obtain defect information present in ROI region；
    S6, prominent defect part；
    S7, structure Bayes classifier；
    S8, on-line checking are simultaneously classified.
  2. 2. a kind of complex-curved defect classification method based on Bayes classifier according to claim 1, its feature exist In：The step S2 image preprocessings include image gray processing and medium filtering.
  3. 3. a kind of complex-curved defect classification method based on Bayes classifier according to claim 1, its feature exist In：The step of including the ROI region of defect part in the step S3 extractions image is as follows：
    S3-1, using iterative method Calculation Estimation threshold value T', be specially：
    S3-1-1, gradation of image average T is asked for by grey level histogram, be designated as initial threshold；
    S3-1-2, use T segmentation figure pictures；Image is divided into all pixels set G1 of the gray value more than T and gray value less than T's All pixels set G2；
    S3-1-3, calculating G1 and the pixel in G2 set average gray value μ 1 and μ 2；
    S3-1-4, using 1/2 (μ 1+ μ 2) the threshold value T's new as one；
    S3-1-5, repeat step S3-1-2~S3-1-4, until the difference of two T' in subsequent iteration is than previously given parameter Be worth it is small untill；
    S3-2, Utilization assessment threshold value T' extract the ROI region for including defect part：
    Analyze image in all pixels point, if pixel gray value is less than threshold value T' in image to be checked, regard it as background dot or Non-detection region point, rejects the pixel；Conversely, being considered as in detection object the point in detection zone, retain the pixel；It is to be checked Altimetric image A (Xi,Yi,Zi) extraction ROI after obtain the new images containing only region to be detectedWherein, Xi,YiFor picture Vegetarian refreshments coordinate, ZiRepresent the pixel gray value.
  4. 4. a kind of complex-curved defect classification method based on Bayes classifier according to claim 1, its feature exist In：The step S4 surface fittings are specially：
    Using least square method structure fitting function Z=F (X, Y)；By imageThe coordinate X of middle pixeli,YiWith Gray value ZiAs fitting function Zi=f (Xi,Yi) sample data, obtain fitting function Z=F (X, Y) and region to be detected and intend Close figure
    The solution of least square fitting is：For the sample point (X on testing imagei,Yi,Zi), look for a letter in function space Number Z=F* (X, Y), make error sum of squares minimum, i.e. the error of fitting result and sample data is minimum, is optimal fitting；Its In, using Euclid norms | | δ | |2As error metrics standard.
  5. 5. a kind of complex-curved defect classification method based on Bayes classifier according to claim 1, its feature exist In：In the step S6, by step S5 difference result using rim detection and morphology operations remove image border and Tiny flaw part, prominent defect part.
  6. 6. a kind of complex-curved defect classification method based on Bayes classifier according to claim 1, its feature exist In：The step S7 structure Bayes classifiers comprise the following steps that：
    S7-1, structure database model；
    S7-2, sample characteristics extraction, draw characteristic vector group H and G；
    S7-3, structure Bayes classifier.
  7. 7. a kind of complex-curved defect classification method based on Bayes classifier according to claim 6, its feature exist In：The step S7-2's comprises the following steps that：
    S7-2-1, obtain positive sample and the gray feature value of negative sample by grey level histogram, the eigenvalue cluster drawn is into an ash Characteristic vector group is spent, is designated as H, the evaluation criterion as gray feature；
    S7-2-2, analyzed by Blob to obtain the geometrical characteristic of positive sample and negative sample, the eigenvalue cluster drawn is into a geometry Characteristic vector group, is designated as G, the evaluation criterion as geometric properties.
  8. 8. a kind of complex-curved defect classification method based on Bayes classifier according to claim 6, its feature exist In：The step S7-3's comprises the following steps that：
    S7-3-1, prepare training sample：
    There to be same characteristic attribute, same category of positive negative sample is divided into same set；Wherein, characteristic attribute is step The gray feature Vector Groups H and geometric properties Vector Groups G that S7-2 is drawn；
    S7-3-2, training grader：
    Calculate three prior probabilities, respectively P (D), P (f | D), P (D | f), generate grader；And then calculated by Bayes' theorem Go out the posterior probability of all kinds of positive negative samples, realize classification.
    Bayes' theorem calculation formula is：
    Wherein, P (f) represents to assume defect f prior probability, and f is a division for assuming set；P (D) represents positive and negative training sample This prior probability；The probability distribution of each characteristic value during P (D | f) is represented per a kind of defect sample, and P (f | D) represent through Bayes The posterior probability for all kinds of defects of hypothesis that classifier calculated obtains；
    S7-3-3, testing classification performance：
    Using the posterior probability of the step S7-3-2 all kinds of defects of classifier calculated for training to obtain, divide corresponding to maximum a posteriori probability Class exports as classification results, by output result and standard results contrast verification, judges whether classification is accurate.
CN201710748823.6A 2017-08-28 2017-08-28 Bayesian classifier-based complex curved surface defect classification method Expired - Fee Related CN107622277B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710748823.6A CN107622277B (en) 2017-08-28 2017-08-28 Bayesian classifier-based complex curved surface defect classification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710748823.6A CN107622277B (en) 2017-08-28 2017-08-28 Bayesian classifier-based complex curved surface defect classification method

Publications (2)

Publication Number Publication Date
CN107622277A true CN107622277A (en) 2018-01-23
CN107622277B CN107622277B (en) 2020-09-22

Family

ID=61089194

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710748823.6A Expired - Fee Related CN107622277B (en) 2017-08-28 2017-08-28 Bayesian classifier-based complex curved surface defect classification method

Country Status (1)

Country Link
CN (1) CN107622277B (en)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108288260A (en) * 2018-01-30 2018-07-17 苏州亚相素自动化科技有限公司 The image pre-processing method of real-time deep or light correction
CN109145977A (en) * 2018-08-15 2019-01-04 河海大学常州校区 A kind of bone damage type identification method based on naive Bayesian
CN110349133A (en) * 2019-06-25 2019-10-18 杭州汇萃智能科技有限公司 Body surface defect inspection method, device
CN111160153A (en) * 2019-12-17 2020-05-15 华南理工大学 Road surface drainage monitoring and evaluating method and system based on image processing
CN111767914A (en) * 2019-04-01 2020-10-13 佳能株式会社 Target object detection device and method, image processing system, and storage medium
WO2021133801A1 (en) * 2019-12-23 2021-07-01 Boon Logic Inc. Product inspection system and method
CN113344042A (en) * 2021-05-21 2021-09-03 北京中科慧眼科技有限公司 Road condition image model training method and system based on driving assistance and intelligent terminal
CN115082424A (en) * 2022-07-19 2022-09-20 苏州鼎纳自动化技术有限公司 3D detection method of liquid crystal display screen
CN115187586A (en) * 2022-09-06 2022-10-14 常州微亿智造科技有限公司 Self-adaptive detection method for workpiece over-cutting and gouging defects
CN116152242A (en) * 2023-04-18 2023-05-23 济南市莱芜区综合检验检测中心 Visual detection system of natural leather defect for basketball

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140270347A1 (en) * 2013-03-13 2014-09-18 Sharp Laboratories Of America, Inc. Hierarchical image classification system
CN104399792A (en) * 2014-11-28 2015-03-11 广东工业大学 Naive Bayes classifier based line heating flame channel point determination method
CN106996935A (en) * 2017-02-27 2017-08-01 华中科技大学 A kind of multi-level fuzzy judgment Fabric Defects Inspection detection method and system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140270347A1 (en) * 2013-03-13 2014-09-18 Sharp Laboratories Of America, Inc. Hierarchical image classification system
CN104399792A (en) * 2014-11-28 2015-03-11 广东工业大学 Naive Bayes classifier based line heating flame channel point determination method
CN106996935A (en) * 2017-02-27 2017-08-01 华中科技大学 A kind of multi-level fuzzy judgment Fabric Defects Inspection detection method and system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
张平等: "基于人工蜂群算法的贝叶斯网络结构学习", 《智能系统学报》 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108288260A (en) * 2018-01-30 2018-07-17 苏州亚相素自动化科技有限公司 The image pre-processing method of real-time deep or light correction
CN109145977A (en) * 2018-08-15 2019-01-04 河海大学常州校区 A kind of bone damage type identification method based on naive Bayesian
CN109145977B (en) * 2018-08-15 2021-12-10 河海大学常州校区 Bone damage type discrimination method based on naive Bayes
CN111767914A (en) * 2019-04-01 2020-10-13 佳能株式会社 Target object detection device and method, image processing system, and storage medium
CN110349133A (en) * 2019-06-25 2019-10-18 杭州汇萃智能科技有限公司 Body surface defect inspection method, device
CN110349133B (en) * 2019-06-25 2021-11-23 杭州汇萃智能科技有限公司 Object surface defect detection method and device
CN111160153B (en) * 2019-12-17 2023-03-28 华南理工大学 Road surface drainage monitoring and evaluating method and system based on image processing
CN111160153A (en) * 2019-12-17 2020-05-15 华南理工大学 Road surface drainage monitoring and evaluating method and system based on image processing
WO2021133801A1 (en) * 2019-12-23 2021-07-01 Boon Logic Inc. Product inspection system and method
CN113344042A (en) * 2021-05-21 2021-09-03 北京中科慧眼科技有限公司 Road condition image model training method and system based on driving assistance and intelligent terminal
CN115082424A (en) * 2022-07-19 2022-09-20 苏州鼎纳自动化技术有限公司 3D detection method of liquid crystal display screen
CN115187586A (en) * 2022-09-06 2022-10-14 常州微亿智造科技有限公司 Self-adaptive detection method for workpiece over-cutting and gouging defects
CN116152242A (en) * 2023-04-18 2023-05-23 济南市莱芜区综合检验检测中心 Visual detection system of natural leather defect for basketball

Also Published As

Publication number Publication date
CN107622277B (en) 2020-09-22

Similar Documents

Publication Publication Date Title
CN107622277A (en) A kind of complex-curved defect classification method based on Bayes classifier
CN107657279B (en) Remote sensing target detection method based on small amount of samples
CN110334706B (en) Image target identification method and device
CN110314854B (en) Workpiece detecting and sorting device and method based on visual robot
CN108898610B (en) Object contour extraction method based on mask-RCNN
CN110543837B (en) Visible light airport airplane detection method based on potential target point
CN108562589B (en) Method for detecting surface defects of magnetic circuit material
CN110163853B (en) Edge defect detection method
CN115082419B (en) Blow-molded luggage production defect detection method
CN105740945B (en) A kind of people counting method based on video analysis
CN113724231B (en) Industrial defect detection method based on semantic segmentation and target detection fusion model
CN111582294B (en) Method for constructing convolutional neural network model for surface defect detection and application thereof
CN114972356B (en) Plastic product surface defect detection and identification method and system
CN104636749A (en) Target object detection method and device
CN111914902B (en) Traditional Chinese medicine identification and surface defect detection method based on deep neural network
CN110728185B (en) Detection method for judging existence of handheld mobile phone conversation behavior of driver
CN114627383A (en) Small sample defect detection method based on metric learning
CN113781585B (en) Online detection method and system for surface defects of additive manufactured parts
CN113221956B (en) Target identification method and device based on improved multi-scale depth model
WO2020119624A1 (en) Class-sensitive edge detection method based on deep learning
CN107729863B (en) Human finger vein recognition method
CN113435460A (en) Method for identifying brilliant particle limestone image
CN113989196A (en) Vision-based earphone silica gel gasket appearance defect detection method
Li et al. Detection of small size defects in belt layer of radial tire based on improved faster r-cnn
CN112862767B (en) Surface defect detection method for solving difficult-to-distinguish unbalanced sample based on metric learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20200922

Termination date: 20210828